Linear Full Duplex System and Method for Acoustic Echo Cancellation

1. A method, comprising: calculating, via an echo estimation module, an echo estimate of an echo due to a far end signal associated with a communication; obtaining a residual echo by subtracting, via an echo subtraction module, the echo estimate from a near end signal associated with the communication; decorrelating, using a decorrelation module, the far end signal and the residual echo, wherein output of the decorrelation module is used for adaptations of an adaptive filter module; detecting and managing, via a dual double talk detector, double talk associated with the communication, comprising: computing, via a power based detector, a maximum power of the far end signal over a double talk detection window; and computing, via a correlation based detector, correlation between the echo estimate from the echo estimation module and the near end signal and between the residual echo and the near end signal; determining, via an acoustic echo canceller controller state machine, presence of the double talk based on the computed power and the computed correlations; and providing status information associated with the double talk to the acoustic echo canceller controller state machine.

2. The method of claim 1 , further comprising: detecting, via a saturation detector, saturation associated with the near end signal of the communication; and bypassing adaptation of the adaptive filter module update upon detection of the saturation.

3. The method of claim 1 , further comprising: generating, via a comfort noise generator, a spectrally-matched comfort noise; and mixing, via the comfort noise generator, the comfort noise with the residual echo.

4. A machine-readable medium embodying a set of instructions that, when executed by a machine, causes the machine to: calculate, via an echo estimation module, an echo estimate of an echo due to a far end signal associated with a communication; obtain a residual echo by subtracting, via an echo subtraction module, the echo estimate from a near end signal associated with the communication; decorrelate, using a decorrelation module, the far end signal and the residual echo, wherein output of the decorrelation module is used for adaptations of an adaptive filter module; detect and manage, via a dual double talk detector, double talk associated with the communication, the step of detecting and managing comprising: computing, via a power based detector, a maximum power of the far end signal over a double talk detection window; and computing, via a correlation based detector, correlation between the echo estimate from the echo estimation module and the near end signal and between the residual echo and the near end signal; determine, via an acoustic echo canceller controller state machine, presence of the double talk based on the computed power and the computed correlations; and provide status information associated with the double talk to the acoustic echo canceller controller state machine.

5. A system, comprising: an echo estimation module to calculate an echo estimate of an echo due to a far end signal associated with a communication; an echo subtraction module to obtain a residual echo by subtracting the echo estimate from a near end signal associated with the communication; a decorrelation module to decorrelate the far end signal and the residual echo; an adaptive filter module to receive output of the decorrelation module for performing adaptations of the adaptive filter module; a dual double talk detector to detect and manage double talk associated with the communication, wherein the dual double talk detector comprises: a power based detector to compute a maximum power of the far end signal over a double talk detection window; and a correlation based detector to compute correlation between the echo estimate from the echo estimation module and the near end signal and between the residual echo and the near end signal; and an acoustic echo canceller controller state machine to determine presence of the double talk based on the computed power and the computed correlations and to receive status information associated with the double talk.

6. The system of claim 5 , further comprising: a saturation detector to detect saturation associated with the near end signal of the communication and to bypass adaptation of the adaptive filter module updates upon detection of the saturation.

7. The system of claim 5 , further comprising: a comfort noise generator to generate a spectrally-matched comfort noise and to mix the comfort noise with the residual echo.

8. The system of claim 5 , further comprising: a steady state detector state machine to detect a state selected from the group consisting essentially of a steady state and a transient state associated with the communication; during the steady state, a dynamic activity detection and switching state machine to dynamically detect at least one of internal substates associated with the communication and to dynamically detect a transition between the internal substates and associated configurations for the adaptive filter module, wherein the internal substates include at least one element selected from the group consisting essentially of a receive only state, a transmit only state, a receive-to-transmit state, a transmit-to-receive state, and an idle state; and the acoustic echo canceller controller state machine to refine the transition between internal substates using an output of the dual double talk detector, to manage transition speeds across different internal substates based on a current substate and a substate to which transition is happening, and to override a transition hangover using the double talk status information and cause a faster transition from a single talk state to a double talk substate.

9. The system of claim 8 , wherein the configurations for the adaptive filter module comprises adapting with a normal rate of adaptation during a steady state and in the receive only state; adapting with a slow rate of adaptation during a steady state and in a substate selected from the group consisting essentially of the receive-to-transmit state and the transmit-to-receive state; bypassing adaptation during a steady state and in a substate selected from the group consisting essentially of the transmit only state and the idle state; adapting with a normal rate of adaptation during single talk during a transition to and from the receive only state and bypassing the adaptation during double talk during a transition to and a transition from the receive only state.

10. The method of claim 1 , further comprising: detecting, via a steady state detector state machine, a state selected from the group consisting essentially of a steady state and a transient state associated with the communication; during the steady state, dynamically detecting, via a dynamic activity detection and switching state machine, at least one of internal substates associated with the communication, wherein the internal substates include at least one element selected from the group consisting essentially of a receive only state, a transmit only state, a receive-to-transmit state, a transmit-to-receive state, and an idle state; dynamically detecting, via the dynamic activity detection and switching state machine, a transition between the internal substates and associated configurations for the adaptive filter module; refining, via an acoustic echo canceller controller state machine, the transition between internal substates using an output of the dual double talk detector; managing transition speeds across different internal substates based on a current substate and a substate to which transition is happening; and using the double talk status to override, via the acoustic echo canceller controller state machine, a transition hangover and cause a faster transition from a single talk state to a double talk sub state.

11. The method of claim 10 , wherein the configurations for the adaptive filter module comprise: adapting with a normal rate of adaptation during a steady state and in the receive only state; adapting with a slow rate of adaptation during a steady state and in a substate selected from the group consisting essentially of the receive-to-transmit state and the transmit-to-receive state; bypassing adaptation during a steady state and in a substate selected from the group consisting essentially of the transmit only state and the idle state; adapting with a normal rate of adaptation during single talk during a transition to and from the receive only state; and bypassing adaptation during double talk during a transition to and a transition from the receive only state.

12. The method of claim 10 , further comprising: defining maximum and minimum values of gain (Gres) that linearly attenuates the residual echo without using a non-linear processing element; and managing, as part of the dynamic activity detection and switching state machine, changes to the values of the gain (Gres) as transitions happen from one substate to another.

13. The method of claim 10 , further comprising: backing up, via a backup and restore module, filter coefficients associated with the echo estimation module based on decision of the dynamic activity detection and switching state machine; and restoring, via the backup and restore module, the filter coefficients for use with the echo estimation module based on the decision of the dynamic activity detection and switching state machine.

14. The system of claim 8 , further comprising: a backup and restore module to back up filter coefficients associated with the echo estimation module based on decision of the dynamic activity detection and switching state machine, and to restore the filter coefficients for use with the echo estimation module based on the decision of the dynamic activity detection and switching state machine.

15. The machine-readable medium of claim 4 , further having instructions to: detect, via a steady state detector state machine, a state selected from the group consisting essentially of a steady state and a transient state associated with the communication; during the steady state, dynamically detect, via a dynamic activity detection and switching state machine, at least one of internal substates associated with the communication, wherein the internal substates include at least one element selected from the group consisting essentially of a receive only state, a transmit only state, a receive-to-transmit state, a transmit-to-receive state, and an idle state; dynamically detect, via the dynamic activity detection and switching state machine, a transition between the internal substates and associated configurations for the adaptive filter module; refine, via an acoustic echo canceller controller state machine, the transition between internal substates using an output of the dual double talk detector; manage transition speeds across different internal substates based on a current substate and a substate to which transition is happening; and use the double talk status to override, via the acoustic echo canceller controller state machine, a transition hangover and cause a faster transition from a single talk state to a double talk sub state.

16. The machine-readable medium of claim 15 , wherein the configurations for the adaptive filter module comprise: adapting with a normal rate of adaptation during a steady state and in the receive only state; adapting with a slow rate of adaptation during a steady state and in a substate selected from the group consisting essentially of the receive-to-transmit state and the transmit-to-receive state; bypassing adaptation during a steady state and in a substate selected from the group consisting essentially of the transmit only state and the idle state; adapting with a normal rate of adaptation during single talk during a transition to and from the receive only state; and bypassing adaptation during double talk during a transition to and a transition from the receive only state.